CIMAT: a transpose SRAM-based compute-in-memory architecture for deep neural network on-chip training.
Hongwu JiangXiaochen PengShanshi HuangShimeng YuPublished in: MEMSYS (2019)
Keyphrases
- network on chip
- network architecture
- associative memory
- power dissipation
- power consumption
- data transfer
- routing algorithm
- multi processor
- data transmission
- random access memory
- real time
- packet switched
- cmos technology
- embedded dram
- network simulator
- memory management
- low power
- processing units
- memory access
- routing protocol
- high speed